Revision as of 09:50, 6 March 2006

Defining New Words

We said earlier that you can add words to the Mops dictionary while building a program. In fact, that is what programming in Mops is all about. Class names, method names, and object names in your program all become part of the dictionary as Mops words. Defining new Mops words also lets you write your own routines and subroutines by defining one short and simple word to take the place of several (or many) other Forth statements. This allows you to factor a program into many small and manageable pieces, an important programming concept and technique.

The first definition exercise will be to define a new word that takes care of the symbols in a simple addition problem. The new word is "ADD", although you could choose any word not already in the Mops dictionary.

Note: Actually, there is nothing stopping you from defining words (including classes) with names that already exist in the Mops dictionary. In fact, you may have discovered that Mops will not complain if you do this! While this is an intentional part of Mops' design, it is generally not a good idea to redefine words already in the dictionary unless you have a good idea of what you are trying to accomplish.

The safest way to check that the name of a new word you want to define is not already in the dictionary, is to issue the "tick" command with the name you want to test for. In Mops, as in most Forths, a tick is an apostrophe ('). By typing apostrophe, space, and the name of the word you're testing for, Mops searches the dictionary for an occurrence of that word. If the word is in the dictionary, tick will leave a number on the stack, which is the location in memory of the word's definition. But if the word is not in the dictionary, the message undefined word appears on the screen, which in this case means you're in the clear to define a word with that name.

' window .
5241310 
' twindow .
 
Error # -13  undefined word
' twindow .
         ^
Current object:  TW    class:  MLTEFWIND

You define a new Mops word by typing a colon (:), a space, the name of the new word, one or more spaces or tabs, the sequence of values and/or commands to be performed when you use that new word, and then a final semicolon (;), indicating the end of your new definition. This kind of Mops definition is called, aptly enough, a colon definition. (Note that :CLASS and :M are other, more special purpose defining words.)

Here's an example that defines the new word, ADD, which will perform the addition of two numbers on the stack, display the results, and move the Mops prompt to the left margin of the next line. (We include the stack comment here for completeness and as an example of good practice; you need not type it in.) Recall that each line you type must be terminated by the ENTER key:

 : ADD  ( n1 n2 --  )  + . cr ;
 

The + operation expects to find two numbers on the stack, so to use your new word you would type two numbers (which go onto the stack) and then the new word:

2 6 add
8

A good exercise at this point would be to define new words to perform some of the other arithmetic operations.

The Return Stack

As we have seen, a Mops program basically consists of a sequence of words and messages sent to objects (which cause a method to be executed). The definitions of these words and methods can contain many other words and messages. If you think about what must happen when Mops is executing one definition or method, you can see that when it has to go and execute other words or methods, it will then need to come back to where it was. It needs to mark its place in some way. The way this is handled is with a second stack, similar in operation to the parameter stack, called the return stack. When Mops has to execute something somewhere else, it saves its current position as an address on the return stack. In that other place, if it has to go yet somewhere else, it pushes the new address on to the return stack as well. This is how words or methods can call other words or methods which can in turn do the same, down to a great depth. And by using a second stack, all these return addresses on the return stack don't interfere with items on the parameter stack.

Note: Mops puts some other items besides return addresses on the return stack, and so can you. However, properly-factored object-oriented programs should generally not require direct manipulation of the return stack, so we will not cover how to do this in this Tutorial.

Normally you won't need to worry about what's going on with the return stack, but when there is an error however, it's usually very useful to know what the program was executing when the error occurred. Mops will try to help you do this, however, by reporting errors as we saw in the error message above. Let's look at it again:

' twindow .

Error # -13  undefined word
' twindow .

Current object:  TW    class:  MLTEFWIND

The first line of the error message tells the nature of the error that occurred, in this case, it was an undefined word error. The second line shows the program statement that produced the error. The third line uses a caret (^) to indicate where in the program statement (shown on the second line) that Mops discovered the error. The fourth line cites the object (and its class) where the error occurred, which in this case is object TW (located in the source file zFrontEnd) in 'System source'), which is part of the interpretive behaviors of the Mops window.

We cannot display the name of the method that the error occurred because method names are not stored in a readable form. However, seeing the program statements where the words occurred is usually enough to track down the location of an error.

Named Input Parameters

Mops can make things a little easier for you by reducing concern about the order in which data are stored on, and recalled from, the parameter stack. Whenever you define a new Mops word, Mops lets you assign names to the parameters that are passed to it. After that, you needn't worry about the stack or the order of the data. When you need a datum for an operation, simply refer to it by the name you have assigned to it.

As an example, we will use the multiply-then-divide problem described in [Chapter3.html Lesson 3]. If you recall, the operation was presented as:

  5 * 12 * 50
 -------------
      40
 

To calculate this without named input parameters, just as we did in that lesson, you had to multiply the three numbers in the numerator, and then place the denominator on the stack before dividing. See how this is simplified in a definition that performs the math with named input parameters:

 : FORMULA  { denom n1 n2 n3 -- solution }
            n1 n2 n3 * * denom /  ;
 

The magic of named input parameters takes place inside the braces ({ and }, also called curly brackets). The syntax is deliberately similar to a stack comment, because it is in fact a kind of stack description. So in this case, whenever the word FORMULA is executed, like this:

40 5 12 50 formula .
75

the first thing that happens is that the values are taken (removed) from the stack and put in a special area of memory where they are associated with the names in the curly brackets in the same order as they were put on the stack. Once that happens, their order is no longer important. Their names are used to fill in the values in the calculation.

But note that the ˜solution' parameter is actually a comment”anything between the -- and the } is treated as a comment. You should use this comment area to indicate what your definition leaves on the stack, exactly as in a normal stack comment.

It is important to bear in mind that the names and values you assign to named input parameters are valid only within their own colon definition. You could use the same names with the same or different values in other colon definitions without any interference.

Named input parameters become very powerful in the way you can adjust their values in the course of a colon definition. Consider for example, this formula:

a^2 + b^2

Since the computer can compute only one square at a time, it needs to hold the result of one square while it calculates the second before it can add the two squares. A definition for a word equivalent to this formula would be:

 : FORMULA1  { a  b -- solution }
             a  a  *  -> a
             b  b  *
             a  +    .  cr  ;
 

The "arrow" (gazinta) operation, ->, stores the value currently on the stack (the result of a-squared) into the named parameter, a. This overwrites the original value in a, which came from the stack in the opening instant of this definition's execution. Near the end of execution, a is recalled to be added to the results of b times b. To solve the same formula without named input parameters would require several stack manipulations that sometimes trip up even the experts.

Incidentally, there are other operations you can perform on a number stored as a named input parameter. You can add a number to what is there, or subtract a number from what is there, with the ++> and --> operations. For example, doing

10 ++> denom

inside a colon definition adds ten to a value stored in a a hypothetical named input parameter named denom.

Local Variables

While we're at it, we'll also introduce you to a similar concept, called local variables. They too, appear inside curly brackets within a colon definition, but instead let you assign names to intermediate results that can occur inside such a definition. Local variables are preceded by a backslash. Take this formula, for instance:

  ( a + b - 3c )
 ----------------
    ( b + 2c )
 

The word definition would be:

 : FORMULA2  { a  b  c  \  num den -- result }
             a  b  +  3  c  *  -  -> num
             2  c  *  b  +        -> den
             num  den  /  ;
 

In this example, a, b, and c in the curly brackets are named input parameters that take on the values on the stack. Names after the backslash (\) but before the "--"), are local variables that will be called into action within the definition. In the example, the numerator and denominator are calculated separately and stored (->) in their respective local variables. Then, the local variables are recalled in the proper order for the division operation to produce the result.

Note: You do not need to "initialize" a local variable before using it. You can rely on local variables being initialized to zero at the beginning of a definition.